. lots of questions after today's interview. The problem with Southwest sounds like a typical, decades-old linear programming or operations research problem where you have a bunch of nodes and edges, subject to certain constraints, and you must...
Conversation
...run the model to find the best solution (minimizing cost, for instance, in terms of $ and/or time). There are textbooks written about this, e.g.:
ftp.idu.ac.id/wp-content/upl
Replying to
Southwest seems to have folks working on this (search for "southwest operations research" on LinkedIn). The question is why haven't they improved on SkySolver and was SkySolver the problem, or their other processes? See e.g. this post on the Tableau blog:
4
Why haven't they added constraints to avoid the ridiculous outcomes like circular deadheading? These optimization algorithms are many decades old and today we certainly have the compute to solve these problems and I would imagine...
1
solve them in a realtime streaming way (rather than batched). Is *that* the primary problem? I doubt the size of Southwest's fleet is the problem?
Quote Tweet
Wow... why isn't this reason for Southwest's board of directors to fire the CEO and CTO?
SkySolver seems to be trademarked/owned by General Electric. Never a good sign.
Show this thread
1
Finally, SW threw out a huge number to "fix" the problem but it doesn't seem to pass a basic smell test? We can add a lot of SaaS tools to this developer bill and not get anywhere close to $1 billion. Do they know what they're doing?
Quote Tweet
SW just said it'll cost $1 billion to fix. That sounds... insane?
If you had 100 excellent developers at $300k/ea it's $30m
What's the other 97% for?
Show this thread
1
It sounds like a case of the "left hand doesn't know what the right hand is doing" if these siloed software solutions don't talk to each other
